System Level Design as Applied to CMU Wearable Computers ASIM SMAILAGIC DAN SIEWIOREK Institute for Complex Engineered Systems, Carnegie Mellon University, Pittsburgh, PA 15213 Abstract: The paper describes a system level design approach to the wearable computers project at Carnegie Mellon University (CMU). The project is an unique example of a cross-disciplinary effort, drawing students from mechanical engineering, electrical and computer engineering, computer science, and industrial design. Over the last six and half years that the course has been taught, teams of undergraduate and graduate students have designed and fabricated sixteen new generations of wearable computers, using an evolving artifact-specific, multidisciplinary design methodology. The complexity of their architectures has increased by a factor of over 200, and the complexity of the application has also increased significantly. We introduce a metric to compare wearable computers and show that their performances have increased by several orders of magnitude. A system-level approach to power / performance optimization is going to be a crucial catalyst for making wearable computers an everyday tool for the general public. 1. Introduction of electronics. Thus iterative design and user evaluation made possible by a rapid design/prototyping methodol- Products of the future will be technologically more ogy is essential for quick definition of this new class of sophisticated, highly customized, produced in small computers. We have chosen to create a course centered on batches, and brought to market quickly. These products wearable computers because a majority of engineering de- will require an approach which integrates the application, sign projects require the resolution of multi-designer, con- artifacts, the CAD environment, and physical prototyping. current, highly-constrained and conflicting constraints. These trends are imposing specific requirements on the In the long term, it is clear that a system-level university engineering education and expertise that gradu- approach to power / performance optimization will be a ated students bring to industry. Carnegie Mellon’s inno- crucial catalyst for making wearable computers an every- vative Wearable Computer Course has been offered over day tool for the general public. Energy efficiency must be the last six and a half years. The students in this considered in all phases of system design. multidisciplinary project course design, assemble and fab- Other examples of system design projects include: ricate a new generation of wearable computers for a spe- the Infopad project at UC Berkeley [5], the Wireless Mo- cific client each term. The course builds upon itself in that dem project at UCLA [6], and the Broadband Wired and a history of the design process, technical decisions and Wireless Modem project at IMEC [7]. These projects are improvements can be maintained for different artifact gen- system design efforts that include a multi-disciplinary re- eration. search with an intense interaction with the system indus- Throughout the design of the sixteen generations try. of wearable computers, an interdisciplinary concurrent de- sign methodology (ICDM) has evolved [1],[2],[3],[4]. As a 2. Design Methodology result of the ICDM methodology, we have achieved a four month design cycle for each new generation of wearable The multidisciplinary design evolves in parallel, computers. The cycle time of the new products is ideally ensuring a strong crosstalk among the different discipline suited to the academic semester. The goal of the design teams. The collaboration is attained through a framework methodology is to allow as much concurrency as possible of compatibility between interdisciplinary design tools and in the design process. Concurrency is sought in both time agreement on a design language and representations so and resources. Time is divided into phases. Activities within that the latest information from each discipline is always a phase proceed in parallel, but are synchronized at phase available and understandable to the others. To provide boundaries. flexibility to the decision process during the initial design Since wearable computers represent a new para- stages, each discipline formulates the problem from its point digm in computing, there is no consensus on the mechani- of view in terms of design goals instead of design con- cal/software human computer interface or the capabilities straints. Resources consist of personnel, hardware plat- available for the hardware. The hardware must merge with forms, and communications. Personnel resources are dy- the software design so that adequate resources are avail- namically allocated to groups which focus on specific prob- able for the necessary functions of the software and so that lems. Communications allow design groups and individu- software drivers are available for the hardware. The soft- als to communicate between the synchronization points. ware design must merge with the user interface design so Table 1 depicts the evolution of the ICDM that input/output can be performed and so that the user through the first four generations of wearable computers. has available the functions necessary to perform their task. The first column in Table 1 represents the steps in the prod- Finally, the user interface design must merge with the me- uct cycle from conception through manufacturing that deal chanical/industrial design to capture the interactions be- with the artifact. As the methodology evolved through suc- tween the system and the user. cessive generations of more capable wearable systems, more sub-phases were added as to refine the original phases. 3. Wearable Computers 2.1. Conceptual Product. Over the past 20 years, there has been steady progress toward making computers more mobile. This has During the conceptualization stage, the been driven by both higher processor integration levels multidisciplinary design team establishes a common vi- and the development of high-performance batteries. As sion of the end product. This vision provides a consistent information becomes increasingly available through wire- set of design goals for all disciplines to maintain through- less technology and the Internet, conventional mobile plat- out the product development cycle. Without a common forms (such as laptops) that are heavy and impede mobil- understanding or vision between design groups and their ity quickly become less appealing. members, each would be forced to rely on their own set of The Wearable Computers project at Carnegie assumptions and criteria based on only a single view of the Mellon University is defining the future for not only com- product. As experience was gained with deployment of puting technologies, but also for the sue of computers in wearable computers, the key role of users became appar- daily activities. The goal of this project is to develop a ent. new class of computing systems with a small footprint that can be carried or worn by a human, and be able to interact 2.2 Configurational Design with computer-augmented environments. By rapid prototyping and fast development of new artifacts and During the first two generations, the systems were concepts, CMU has established the new paradigm of wear- derived directly from the feasible technology. Subsequently able computers. The sixteen generations of wearable com- the product configuration became a critical phase. It re- puters have been designed and built over the last six and a sults in system architecture and subsystem specification. half years, and most of them field tested as well. Several of them relevant to this paper are summarized below. The 2.3. Detailed design. first system was a technology exploration vehicle, seek- ing lightweight, low-energy solutions to mobile environ- The detailed design phase is traditionally defined ments with 1) long range and 2) short range, dense infra- by a well developed methodology and a rich set of CAD structure. The following two systems are pentium-based tools. A detailed design of each subsystem is performed, wearable computers in very small packaging. The next with particular attention to maintaining the interface speci- system represents a family of wearable computers dedi- fications as defined in the Product Design Specification. cated to speech translation applications. The last one is a very low power consumption StrongARM based wearable 2.4. Manufacturing computer providing high processing capabilities. • ISAAC (Integrated Speech - Activated Application The manufacturing phase uses a combination of Control) is a technology feasibility prototype which electronic and mechanical job-shops and on-campus rapid explores capabilities in a low weight, low energy area, prototyping facilities. After detailed design is completed, assuming communication over short distances. ISAAC each group implements their subsystem using the acquired employs speech for communication with the infra- technology. As implementation progresses, the subsystems structure and special rooms around campus. A wire- are demonstrated at various stages of development. less microphone transmits analog speech to a speech System design and engineering is performed by recognition system on a base computer. The base the class as a whole and then the various disciplines per- computer can control devices through infrared repeat- form detailed design and implementation. During the whole ers and communicates with the user via an infrared process, the four disciplines interact along well defined wireless